[rrd-users] RRD consolidation and step questions

Marc MERLIN marc_rrd at merlins.org
Sun Oct 18 17:09:53 CEST 2009


I'm looking at graphing my house power use and generation from cacti.

I don't have a lot of data and I do care about keeping high resolution 
over time, so I've configured cacti to have the following RRA levels:

				step	rows		timespan
Hourly (1 Minute Average)   	1  	525600  	14400  		1Y/3H
Daily (5 Minute Average) 	5 	3153600  	86400		30Y/24H
Weekly (30 Minute Average) 	30 	525600		604800 		30Y/7d
Monthly (2 Hour Average) 	120 	131400 		2678400 	30Y/31d
Yearly (1 Day Average) 		1440 	10950 		31536000	30Y/382d

The questions I haven't yet answered by reading the docs are:

1) If I pick the 5m avg, since I have enough samples for 30 years,
do I get any benefit whatsoever from gathering the other ones
(daily/weekly/monthly/yearly) with AVERAGE and/or MAX, or is it indeed
totally pointless since cacti/rrdtool graph will calculate lower res data
if I make it print weekly/monthly graphs?

If I have
/usr/bin/rrdtool create /var/lib/cacti/rra/house_20.rrd --step 120  \
DS:HouseNoAC:ABSOLUTE:600:-1000000000:1000000000 \
DS:House:ABSOLUTE:600:-1000000000:1000000000 \
DS:PGE:ABSOLUTE:600:-1000000000:1000000000 \
DS:AC:ABSOLUTE:600:-1000000000:1000000000 \
DS:PV:ABSOLUTE:600:-1000000000:1000000000 \
RRA:AVERAGE:0.5:1:525600 \
RRA:AVERAGE:0.5:24:3153600 \
RRA:MAX:0.5:1:525600 \
RRA:MAX:0.5:24:3153600 \

2) where is AVERAGE/MAX applied? 
- when 5 data points are consolidated from the 1mn to the 5mn storage?
- or is this done at data sample time with the value in step?
  If so, is it indeed pointless to have both AVG and MAX if your step is 1?
  (or if I want to graph MAX later in cacti, but I only have AVG for my
  1mn resolution data, it won't work?)

3) My data sources only gets updates every 2mn (--step 120). Currently I
have cacti configured to use the 1mn and 5mn RRAs. However,

3a) the 1mn RRA only gets run every other minute. Will that upset things, or rrdtool
obviously does notice that the samples are farther apart and makes it up.

3b) the 5mn RRA only sees 2.5 samples instead of 5 (step). What happens there?
Is rrdtool ok enough with not having as many data samples as the step value in
the RAA and compensates happily too, or do bad things happen?

3c) am I better off creating and using a separate 5mn/2 step RRA where I give it
a step of 2 instead of 5 or using my current 5mn/5 step RRA will work?

3d) more generally, what happens when --step and the step in the RRA are mismatched?
If I get data every minute later, is it easy to change --step 120 to 
--step 60 somehow since my RRAs already have room for data every minute?

Thanks much,
"A mouse is a device used to point at the xterm you want to type in" - A.S.R.
Microsoft is to operating systems & security ....
                                      .... what McDonalds is to gourmet cooking
Home page: http://marc.merlins.org/  

More information about the rrd-users mailing list